The MPEG-2 standard addresses the combining of one or more elementary streams of video audio and other data into single or multiple streams that are suitable for storage or transmission. In very general terms, the MPEG-2 standard for transmitting digital video, the associated audio and other information involves the following three steps. In the first step, a digital video signal (from a digital camera or from an analog to digital converter) is compressed by analyzing and encoding the signal using spatial and temporal redundancy. Spatial redundancy refers to the redundant information inside one video frame while temporal redundancy refers to the redundant information between consecutive frames. This process generates: Intra-frames (I-frames), which contain all of the information in an entire image; Predicted frames (P-Frames), which have some compression as they are predicted based on past I-frames and/or other P-frames; and Bi-directionally predicted frames (B-frames), which are the most compressed images as they are predicted from past and future I-Frames and P-Frames. In the second step carried out concurrently with the first step, an audio signal is compressed by removing low-power tones adjacent high-power tones. Removal of these tones does not affect the signal, because the high-power tones tend to mask the lower-power tones, making them inaudible to the human ear. In the final third step, the compressed, video signals audio signals and related time stamps of those signals are assembled into packets and inserted into a Packetized Elementary Stream (PES). Each packet in a packetized elementary stream contains overhead information such as a start code, stream ID, packet length, optional packetized elementary stream header and stuffing bytes, in addition to the actual packet bytes of video and audio data.
To facilitate the multiplexing together of several streams of packetized elementary streams of different types of data, a Programme Specific Information (PSI) table is also created, which includes a series of tables to reassemble specific packetized elementary stream within multiple channels of packetized elementary streams. The packetized elementary stream and the program specific information provide the basis for a Transport Stream (TS) of packetized elementary stream and program specific information packets.
Of particular interest to the invention disclosed herein is the transport stream as defined in Annex D of the “ATSC Digital Television Standard” published by the Advanced Television Systems Committee ATSC) in September 1995 as its document A/53. This standard defines the broadcasting of digital television (DTV) signals within the United States of America and is referred to in this specification simply as “A/53”. Annex D specifies that the original data transport stream is composed of 187-byte packets of data corresponding to MPEG-2 packets without their initial sync bytes. Annex D specifies that data are to be randomized by being exclusive-ORed the a specific 216-bit maximal length pseudo-random binary sequence (PRBS) which is initialized at the beginning of each data field. Annex D specifies (207, 187) Reed-Solomon forward-error-correction (R-S FEC) coding of packets of randomized data followed by convolutional interleaving. The convolutional interleaving prescribed by A/53 provides error correction capability for continuous burst noise up to 193 microseconds (2070 symbol epochs) in duration. The convolutionally interleaved data with R-S FEC coding are subsequently trellis coded to ⅔ original code rate and mapped into eight-level digital symbols. The symbols are parsed into 828-symbol sequences.
Annex D specifies that the data frame shall be composed of two data fields, each data field composed of 313 data segments, and each data segment composed of 832 symbols. Annex D specifies that each data segment shall begin with a 4-symbol data-segment-synchronization (DSS) sequence. Annex D specifies that the initial data segment of each data field shall contain a data-field-synchronization (DFS) signal following the 4-symbol DSS sequence therein. The DSS and DFS signals are composed of symbols with +5 or −5 modulation signal values. The 2nd through 313th data segments each conclude with a respective one of the trellis-coded 828-symbol sequences, the convolutional interleaving of which sequences extends to a depth of 52 data segments. The digital symbols are transmitted by eight-level modulation with +7, +5, +3, +1, −1, −3, −5 and −7 modulation signal values. Owing to the A/53 baseband DTV signal being transmitted via vestigial-sideband suppressed-carrier amplitude modulation of a radio-frequency carrier, this eight-level modulation signal is referred to as trellis-coded 8VSB signal. These transmissions are accompanied by a pilot carrier of the same frequency as the suppressed carrier and of an amplitude corresponding to modulation value of +1.25.
The fifth through 515th symbols in the initial data segment of each data field are a specified PN511 sequence—i. e., a pseudo-random noise sequence composed of 511 symbols capable of being rendered as +5 or −5 modulation signal values. The 516th through 704th symbols in the initial data segment of each data field are a triple-PN63 sequence. The middle PN63 sequence is inverted in sense of polarity every other data field. The 705th through 728th symbols in the initial data segment of each data field contain a VSB mode code specifying the nature of the vestigial-sideband (VSB) signal being transmitted. The remaining 104 symbols in the initial data segment of each data field are reserved, with the last twelve of these symbols being a precode signal that repeats the last twelve symbols of the data in the last data segment of the previous data field. A/53 specifies such precode signal to implement trellis coding and decoding procedures being resumed in the second data segment of each field proceeding from where those procedures left off processing the data in the preceding data field.
The 8VSB transmissions have a 10.76 million bits per second baud rate to it within a 6-megahertz-wide broadcast television channel, and the effective payload is 19.3 million bits per second (Mbps). In an additive-white Gaussian noise (AWGN) channel a perfect receiver will require at least a 14.9 dB signal-to-nose ratio (SNR) in order to keep errors below a threshold of visibility (TOV) defined as 1.93 data segment errors per 10,000 data segments, supposing 8VSB signals are broadcast.
After the “ATSC Digital Television Standard” was established in 1995, reception of terrestrial broadcast DTV signals proved to be problematic, particularly if indoor antennas were used. In early 2000 ATSC made an industry-wide call for experts in terrestrial broadcast transmission and reception to join a Task Force on RF System Performance for studying problems with adequate reception and suggesting possible solutions to those problems. By the end of 2000 or so there was general consensus that, besides problems with equalization of the reception channel, there was a need to make the 8VSB signal more robust, if it were to be successfully received during noise reception conditions. On 26 Jan. 2001 the ATSC Specialist Group on RF Transmission (T3/S9) issued a “Request for Proposal for Potential Revisions to ATSC Standards in the Area of Transmission Specifications”. This RFP concerning how to improve the performance of 8VSB was directed to the DTV industry, universities and other parties interested in the problem. The widely distributed T3/S9 RFP specifies backward-compatible improvement of fixed and indoor 8-VSB terrestrial DTV service to be of top priority. The requirement for backward-compatibility with legacy DTV receivers means, among other things, that the trellis coding specified in A/53 must be maintained throughout data fields.
A general approach to making 8VSB signal more robust is to increase the amount of forward-error-correction-coding. Zenith Electronics and ATI proposed the application of preliminary additional trellis coding to data before the trellis coding specified by A/53. Legacy DTV receivers already in the field are incapable of receiving the data with the additional trellis coding, however.
The general concept that FEC coding can be contained in data packets that do not contain payload data and that are separate from data packets that do contain payload is found in U.S. Pat. No. 6,430,159 the specification and drawing of which are incorporated herein by reference. U.S. Pat. No. 6,430,159 titled “Forward error correction at MPEG-2 transport stream layer” issued Aug. 6, 2002 to Xiang Wan and Marc H. Morin. Wan and Morin sought to provide a system and method to correct an MPEG-2 transport stream that could be used in any one of the digital video broadcast formats, without the need for FEC decoders which were specific to the particular format. Another objective of the U.S. Pat. No. 6,430,159 invention was to avoid appending FEC coding to the end of each packet, in effect adding another layer to the protocol stack. Such a new layer is specific to the transmission architecture and not subject to the MPEG-2 standard, so a broadcaster has to rely upon each intended receiver having a symmetric FEC decoder for the transmitted signal to be received. However, the satellite broadcast industry the cablecasting industry and the terrestrial broadcast industry embraced the practice of inserting the original transport stream into a forward-error-correction encoder and broadcasting the resulting signal over their respective broadcast medium to receivers. The various receivers for satellite broadcast, cablecasting and terrestrial broadcast systems recover MPEG-2-compliant transport streams from received signals, using FEC-decoders specific to the various systems and symmetric with the FEC encoders employed in these various systems.
A. L. R. Limberg ran across U.S. Pat. No. 6,430,159 during a comprehensive review of DTV receiver practice he conducted in 2002 when working on the revision of the “Guide to the Use of the ATSC Digital Television Standard” published in October 1995 as ATSC Document A/54. Limberg perceived that the Wan and Morin concept still had practical utility, even though (207, 187) R-S FEC coding was appended to data segments of 8VSB DTV broadcast signal, employing the sort of practice Wan and Morin had sought to avoid by their invention. Limberg perceived that transverse Reed-Solomon forward-error-correction coding facilitates additional error-correction coding being time-division multiplexed with A/53 data segments in such a way that DTV receivers already in the field can still receive the A53 data segments. Limberg understood that the Wan and Morin concept was the key to solving the problem of making the DTV signals more robust without making those signals impossible to be received by DTV receivers already in the field. This was the problem that had stumped experts in DTV system design for two years or more despite T3/S9 having focused industry effort on solving this problem. Limberg discerned that transverse R-S FEC coding was orthogonal to “lateral” (207, 187) R-S FEC coding prescribed by A/53 and combined therewith to provide two-dimensional R-S FEC coding.
An alternative approach to making 8VSB signal more robust is to restrict the symbol alphabet in such a way that symbol decoding procedures are less susceptible of error. For example, a set of limited-alphabet 8VSB symbols that map data into just +7, +5, 5 and −7 modulation signal values was proposed by Philips Research responsive to the T3/S9 RFP. This limited-alphabet signal is referred to as “pseudo-2VSB” or “P-2VSB”, since the polarity of the signal suffices to convey the information in the resulting modulation signal. Using P-2VSB throughout the entire DTV broadcast halves the effective payload to 9.64 million bits per second (Mbps), but this is more than sufficient to transmit a standard-definition television (SDTV) signal. The gap between the least negative normalized modulation level, −5, and the least positive normalized modulation level, +5, is 10. This is five times the gap of 2 between adjacent normalized modulation levels in an 8VSB signal. The 8VSB signal has ⅔ trellis coding, however, which increases its performance capability to be somewhat better than a VSB signal with a gap of 4 between adjacent normalized modulation levels. Accordingly, the SNR required in order to keep error below TOV in an AWGN channel is reduced to 8.5 dB, a reduction of 6.4 dB. That is, about a quarter as much power would be required for satisfactory reception of an AWGN channel, presuming that modulation levels did not have to be decreased to maintain effective radiated power (ERP) levels within current specification. The ERP of the P-2VSB symbols tends to increase respective to conventional trellis-coded 8VSB, because of just the +7, +5, −5 and −7 modulation signal values being used and the +3, +1, −1 and −3 modulation signal values of 8VSB not being used. A 1.5 dB decrease in transmitter ERP is necessary if long sequences of P-2VSB symbols are transmitted. So, if long sequences of P-2VSB symbols are transmitted, increase in service area for the P-2VSB signal is only that which could be achieved with a 4.9 dB increase in the power of a conventional trellis-coded 8VSB signal. Furthermore, service area for the conventional trellis-coded 8VSB signal accompanying the P-2VSB signal is diminished. Consequently, P-2VSB sign as shave been considered only for only a limited number of the data segments in each 313 segment data field.
The Electronics and Telecommunications Research Institute (ETRI) and Chonnnam National University (CNU) Multimedia Communications Laboratory in South Korea proposed another set of restricted-alphabet 8VSB subjects that map data into just +7, +1, −3 and −5 normalized modulation signal values. This type of signal is referred to as “enhanced-4VSB” or “E-4VSB”. The gap between the least negative normalized modulation level, −3, and the least positive normalized modulation level, +1, is 4. This is twice the gap of 2 between adjacent modulation levels in an 8VSB signal, permitting TOV under AWGN conditions to be achieved at significantly poorer SNR than is the case with 8VSB signal. The SNR that E-4VSB requires to keep errors below TOV is higher than that required with pseudo-2VSB modulation.
The ERP of the E-4VSB symbols is purportedly the same as that of conventional trellis-coded 8VSB. Consequently, considerably more data segments in data fields can code E-4VSB modulation than can code pseudo-2VSB signals, presuming that ERP is not to be increased very much. E-4VSB is presumably less likely than pseudo-2VSB to disrupt the operation of legacy receivers, particularly those that rely on symbol averaging to develop automatic gain control signals for controlling the gains of their amplifier stages. The limitation on the number of the data segments in each 313-segment data field that can be E-4VSB signals depends solely on the number of data segments of normal transmissions that must be provided to accommodate legacy receivers. The more asymmetrical symbol constellation benefits symbol synchronization using bright-spectral-line techniques.
Certain modifications of the original data transport stream cause the trellis coding procedure at the transmitter to generate 8VSB symbols with various restrictions of the available symbol alphabet. Each bit in a stream of randomized data can be repeated to generate a modified stream of data supplied to the (207, 187) R-S FEC encoder for example, to cause a pseudo-2VSB signal to be generated by the trellis coding procedure. In other procedures for restricting the symbol alphabet for each symbol epoch, each bit in a stream of randomized data can be followed by an additional bit of prescribed value independent of the bit it follows.
By way of example, ONE can be inserted after each bit in a stream of randomized data to generate a modified stream of data supplied to the (207, 187) R-S FEC encoder. This modified stream of data causes the trellis coding procedure to generate a restricted-alphabet signal which excludes the −7, −5, +1 and +3 symbol values of the full 8VSB alphabet. Pilot carrier energy is increased substantially in the resulting modulation, which makes synchronous demodulation easier in the DTV receiver. The gap between the least negative normalized modulation level, −5, and the least positive-normalized modulation level, +1 is 6 in this restricted-alphabet signal. This gap is three times the gap of 2 between adjacent modulation levels in an 8VSB signal, permitting TOV to be achieved at significantly poorer SNR under AWGN conditions than is the case with 8VSB signal or with E-4VSB signal. Better SNR under AWGN conditions is required to achieve TOV than is the case with pseudo-2VSB. This restricted-alphabet signal has substantially less average power than a pseudo-2VSB signal, but somewhat higher average power than normal 8VSB signal.
By way of counter example, a ZERO can be inserted after each bit in a stream of randomized data to generate a modified stream of data supplied to the (207, 187) R-S FEC encoder. This modified stream of data causes the trellis coding procedure to generate a restricted-alphabet signal which excludes the −3, −1, +5 and +7 symbol values of the full 8VSB alphabet. The gap between the least negative normalized modulation level, −5, and the least positive normalized modulation level, +1, is also 6 in this restricted-alphabet signal. However, this restricted-alphabet signal has somewhat less average power than normal 8VSB signal. A difficult problem with using this restricted-alphabet signal is that the polarity of the pilot signal is reversed in the resulting modulation, which interferes with synchronous demodulation in DTV receivers, particularly legacy ones.
The inventor subsequently discerned that the 8VSB alphabet can be restricted in such a way that, in accordance with a prescribed pattern, a ZERO or a ONE is inserted as an X1 bit after each of the X2 bits in a data segment to be incorporated into a data field for randomization, R-S FEC coding, convolutional interleaving, and trellis coding. If ZEROes and ONEs occur with similar frequency in the pattern, ERP can be kept substantially the same as in an ordinary 8VSB signal described in Annex D of A/53. This requires careful selection of the prescribed pattern of inserting ZEROes and ONEs as X1 bits. If symbols are correctly sampled, the gap between the least negative normalized modulation level and the least positive normalized modulation level is 6 in each symbol of this restricted-alphabet signal. This general type of restricted-alphabet signal, constructed from co-sets of a complete symbol alphabet that occur at prescribed times, is an important aspect of certain of the inventions described in this specification. This general type of restricted-alphabet signal is also useful in applications other than 8VSB DTV broadcasting, being useful in MPSK transmissions by way of example.
Viewed another way, this aspect of the invention concerns time-dependent trellis coding in which different sets of symbols are precluded at prescribed times in order to increase the Hamming distances between possible trellis codes. The symbols that are precluded are determined in advance according to a prescribed pattern, which pattern does not depend on the history of previous symbols. The pattern can be chosen to adjust the ERP of a transmitter such that average power is substantially the same as for symbol coding in which symbols are randomly selected from the full 8VSB symbol alphabet. This time-dependent trellis coding differs from extended trellis coding in which the symbols that are precluded are determined depending on the history of previous symbols. This time-dependent trellis coding is not subject to the tendency towards running error in the decoding of trellis code increasing as the code is extended. Each successive symbol in the time-dependent trellis code exhibits increased Euclidean distance respective to other symbols that could occur during that symbol epoch, so the possibility of error in hard-decision decoding is substantially reduced. This can be used for improving adaptive equalizer convergence during rapidly changing multipath conditions.
U.S. patent application Ser. No. 10/733,645 filed 12 Dec. 2003 for A. L. R. Limberg and titled “Robust Signal Transmissions in Digital Television Broadcasting” describes transverse Reed-Solomon forward-error-correction coding being used to supplement the error correction coding already in the 8VSB data segments. The parity bytes for the transverse Reed-Solomon forward-error-correction coding are arranged in A/53-compliant data segments to be time-division multiplexed with conventional A/53 data segments. The resulting signal is then convolutionally interleaved, trellis coded and mapped into 8VSB symbols per subsections 4.2.4 and 4.2.5 of A/53, Annex D. Patent application serial No. 10/733,645 specifically considers how transverse Reed-Solomon forward-error-correction coding of restricted-alphabet signals can be done. Patent application Ser. No. 10/733,645 discloses a problem that is encountered when one attempts to apply transverse Reed-Solomon forward-error-correction coding to restricted-alphabet signals in which the Z1 bit in a symbol codeword elected for the restricted-alphabet signal cannot be determined independently of the Z0 term. Suppose the parity bytes of the transverse R-S FEC coding were permitted to interleave convolutionally with bytes of such a restricted-alphabet signal. Then, the Z0 bits in the symbol codewords of such a restricted-alphabet signal would depend on the Z1 bits in the symbol codewords of the parity bytes of the transverse R-S FEC coding. However, the Z1 bits in the symbol codewords of the parity bytes of the transverse R-S FEC coding should depend on the Z1 bits of the symbol codewords in the restricted-alphabet signal. This is a situation of trying to “lift oneself by one's own bootstraps”. E-4VSB signal has 001, 010, 100 and 111 symbol codewords that respectively generate −5, −3, +1 and +7 normalized modulation signal values. The Z1 bits in the E-4VSB symbol codewords cannot be determined independently of the Z0 bits, so the E-4VSB signal does not lend itself to transverse R-S FEC coding, at least not readily. Accordingly, there is a need for a type of robust modulation that halves code rate without affecting average ERP, but also better lends itself to transverse R-S FEC coding. U.S. patent application Ser. No. 10/733,645 issued 27 Mar. 2007 as U.S. Pat. No. 7,197,685 titled “Robust Signal Transmissions in Digital Television Broadcasting” and is incorporated by reference herein.
The known types of robust modulation that have code rate, but also lend themselves to transverse R-S FEC coding, are ones with a set of four symbol codewords for which the Z1 bits can be determined independently of the Z0 bits. The Z1 bit repeats the Z2 bit in all 3-bit symbol codewords of pseudo-2VSB signals, so pseudo-2VSB modulation lends itself to transverse R-S FEC coding. So does robust modulation wherein accordance with a prescribed pattern a ZERO or a ONE is inserted as a respective Z1 bit after the Z2 bit in each 3 bit-symbol codeword before it is supplied to a trellis encoder.
A previous practice when including robust transmissions DTV signals has been to confine the robust transmissions to the 184-byte payload portions of data segments. Each data segment containing robust transmission begins with a 3-byte header that causes the data segment to be discarded by legacy 8VSB DTV receivers. Each data segment containing robust transmission concludes with twenty parity bytes of R-S FEC coding. MPEG-2 data packets do not map to an integral number of data segments when such previous practice is followed. Accordingly, such previous practice requires rather elaborate procedures for parsing data packets, especially since data segments associated with robust transmission have to be time-division multiplexed with data segments associated with ordinary HDTV transmission. The procedures for parsing data packets are apt to error during noisy reception.
The inventor prefers a new practice for including robust transmissions in DTV signals. In this preferred practice a data segment containing a 187-byte MPEG-2-compliant data packet and twenty bytes of lateral R-S FEC coding is converted into an integral number of consecutive data segments, such as two, which provides for simple parsing. The consecutive data segments generated by this simple conversion procedure will not be A/53 compliant, but this need not be problematic. Legacy DTV receivers are incapable of usefully receiving restricted-alphabet components of an 8VSB DTV broadcast signal anyway. Accordingly, the data segments including robust transmissions are freed from having to meet certain requirements, insofar as accommodating legacy DTV receives is of concern. These data segments do not each need to include a data packet complying with MPEG-2, and these data segments do not each need to include parity bytes of “lateral” (207, 187) R-S FEC coding as prescribed by A/53. These data segments should be ones that legacy DTV receivers will discard during transport stream de-multiplexing, either because they do not appear to include recognizable PID code or because they are found not to be correctable during R-S FEC decoding procedures. Each data packet that is to be transmitted using a restricted symbol alphabet can be evaluated ahead of time. The evaluations are made to ascertain which data randomization sequences would result in a legacy receiver finding one or both of the data segments derived from that data packet to contain both a valid PID and correctable byte errors. Transmission of the robust-data packet is scheduled in the data field so that each portion of that packet in a respective data segment uses a data randomization sequence that results in byte errors beyond the capability of correction by a standard (207, 187) R-S FEC decoder.
U.S. patent application Ser. 10/885,460 filed 6 Jul. 2004 for A. L. R. Limberg and titled “Reed-Solomon Coding Modifications for Signaling Transmission of Different Types of Data Packets” is incorporated by reference in this application. U.S. patent application Ser. No. 10/885,460 discloses an alternative way to cause the robust-data segments to contain byte errors beyond the capability of correction by a standard (207, 187) R-S FEC decoder. Each segment of robust data that contains byte errors within the capability of correction by a standard (207, 187) R-S FEC decoder is modified before transmission so this is no longer the case. The modification causes shortened R-S coding that is different than normal, so a legacy DTV receiver will find the robust-data segment to contain byte errors beyond the capability of correction by its (207, 187) R-S FEC decoder. A new DTV receiver will undo this modification responsive to a byte errors in a data segment being found to be correctable by a (207, 187) R-S FEC decoder for the shortened coding that is different than normal.
A DTV receiver that is adapted for useful receiving both full-alphabet and restricted-alphabet components of an 8VSB DTV broadcast signal has to have knowledge of when each of these components is being received. This knowledge permits symbol decoding of the restricted-alphabet components to be done in special way that improves the accuracy of symbol decoding decisions. The general procedure in the prior art is for the DTV transmitter to transmit information to the DTV receiver concerning the pattern of data segments recovered from restricted-alphabet components of the 8VSB DTV broadcast signal, which pattern obtains in each data field before being convolutionally interleaved and trellis coded. This information is transmitted in the reserved portion of the initial data segments of data fields, various coding schemes for such information being known. U.S. Pat. No. 6,563,436 titled “KERDOCK CODING AND DECODING SYSTEM FOR MAP DATA” and issued 13 May 2003 to M. Fimoff, R. W. Citta and J. Xia describes one way of doing this, for example. The pattern information is convolutionally interleaved to generate information concerning which symbols of the convolutionally interleaved data field received by the DTV receiver are selected from a restricted alphabet of 8VSB symbols. Certain of the data segments in the de-interleaved field that the de-interleaver generates from trellis coding results are recovered from restricted-alphabet components of the 8VSB DTV broadcast signal. The pattern information available to a TV receiver is used in an additional way in novel DTV receivers described in this specification and its drawing. The pattern information is used to select these data segments for the data compression that converts them to a reduced number of data segments that comply with A/53 standards for data segments recovered from full-alphabet components of the 8VSB DTV broadcast signal.